# If people use a scientific way to predict the weather, why is it sometimes the predictions are wrong?

##### 1 Answer

The number of assumptions is very high.

#### Explanation:

If you have ever heard the term "chaos theory" it does a great deal to answer this question. The main idea behind chaos theory is not actually chaos, just the inability to measure all the variables. It is actually the furthest thing from chaos, it is extreme order. The idea that we can't predict certain events because we don't even know all the variables never mind measuring them all. The underlying idea begin that if we could know what all the variables are, and we could measure them all, we could predict anything with 100% accuracy.

In weather there are often tens to hundreds of kms between reporting sites. In fact there is an area in North America that is hundreds of thousands of square kms with no reporting sites. How do we know what is going on with the air pressure, for example, between these widespread sites? We use computer models.

A computer model essentially averages out the values for numbers in data sparse areas but also in populated areas, since local effects can sometimes only effect a few kms. This gives us a starting point, the now picture of the variables. Since this rounding out is done by a computer, a lot of the data is correct, but some is wrong. The models however, assume that there are no errors.

If the now picture of atmospheric variables is not 100% accurate, even if it is 99% accurate, then subsequent predictions are built on partial errors. So if we are looking at a few hours from the starting point and there was a 1% error rate at the starting point, then a few hours from now there would be 2% error rate. This generally doesn't effect much, but as the computer continues assuming that it is 100% correct, the errors continue to pile up. At a 1 day we might be looking at a 10% error rate, by day three a 30-40% error rate.

The other thing to consider is modernly educated meteorologist are trained more to build models than they are to actually look at data and come up with analyses on their own. Further more, cut backs in atmospheric services around the globe often result in forecasters being responsible for a larger area than they may should have. This results in a "good enough" mentality.

The forecaster is busy so when he looks at the model he doesn't look overly closely at it. He basically says it's good enough and I can fix the forecast later if it is incorrect. It is relatively easy to amend a forecast, and very expensive to hire another forecaster to break up the area of responsibility.